CI optimization playbook: measure the critical path, cut agent wait time ~40%
RishabJainK · x · 2026-11-03
The full CI speedup methodology behind the 36% gate-time reduction:
- Measure what CI waits on: pull 40+ recent runs, record total and per-job times, find the critical path (parallel jobs tie together—speeding a non-slowest job helps nothing), then split the slowest job's time into machine wait / setup (checkout, install, image pull, cache restore) / actual work, and fix the biggest slice.
- Slow tests hide waiting, not work: use per-file test times to find the few files setting each shard's floor—shards can't finish before their slowest file.
Chart shows median gate time dropping 300s→192s via fixes like short waits against instant fakes, shard rebalancing, short-lived build VMs, and an 8s pre-push hook.
Related event: Boardy Cuts CI Time from 40 Minutes to 2 with 20x Speedup Playbook(2 posts)→
More from coding & agent
- Article explores how AI agents can improve their own harness without retraining — dima806_dima · 2026-11-03
- Slop Cannon plugin for Nous Research's Hermes agent renders multi-scene videos with a native Studio UI — Teknium · 2026-11-03
- Honest review of esolang Bend2: no type inference, 9/10, 'still too easy for JS devs' — max_paperclips · 2026-11-03
- Dev claims WebMCP beats browser-use agents for web automation — aniketmaurya · 2026-11-03
- CUA-Master: Nested Dual-Loop Framework Scales Computer-Use Agents via Open-World Recursive Improvements — Xinlei Yu · 2026-11-03
- CrossSWE benchmark shows coding agents struggle with cross-repository development — Silin Chen · 2026-11-03